Ophthalmic research · 2026
Diagnostic Performance of Deep Learning Using Retinal Images for Detection of Chronic Kidney Disease: A Systematic Review and Meta-analysis.
СтатьяОбзорНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Farrokhi Mehrdad, Hoorshad Behnam, Fahimi Reza, Moghadam Fard Atousa, Dalvandi Behnaz, Dalvandi Reza, Esmaielzade Rostami Mahshid, Judi Chelan Reza, Ahmadzadeh Oveis, Salarinejad Sareh, Shemshadigolafzani Ramtin, Alipour Morteza, Rahmani Parham, Lashgari Marjan, Asadalizadeh Mahya
Аннотация
BACKGROUND: Multiple investigations have been conducted to diagnose chronic kidney disease (CKD) from photographs of retina using deep learning models, with a wide range of diagnostic performance, which necessitates the importance of a comprehensive review. Therefore, in this meta-analysis, we estimated the detection performance of deep learning models for the diagnosis of CKD using retinal photographs.
METHODS: MEDLINE, Scopus, and Web of Science were searched until 25 December 2025. Pooled estimation of heterogeneity, diagnostic parameters, and meta-regressions were performed using MetaDisc and STATA Softwares.
RESULTS: The pooled sensitivity, specificity, and diagnostic odds ratio (DOR) were 0.81 (95% CI, 0.80-0.82), 0.64 (95% CI, 0.64-0.65), and 9.15 (95% CI, 5.72-14.62), respectively. The pooled estimates of the positive and negative likelihood ratios were 2.18 (95% CI, 1.66-2.87) and 0.28 (95% CI, 0.20-0.39), respectively. The area under the summary receiver operating characteristic (SROC) curve showed a discriminative performance of 0.77.
CONCLUSION: Deep learning based on retinal images had moderate diagnostic performance for detection of CKD. Relatively high sensitivity and modest specificity suggesting that deep learning is more effective for ruling out CKD than for confirming the diagnosis that deep learning is more effective for ruling out CKD than for confirming the diagnosis. Deep learning-based retinal analysis may therefore serve as a supportive, noninvasive screening or triage tool but should not replace established diagnostic methods for CKD.
METHODS: MEDLINE, Scopus, and Web of Science were searched until 25 December 2025. Pooled estimation of heterogeneity, diagnostic parameters, and meta-regressions were performed using MetaDisc and STATA Softwares.
RESULTS: The pooled sensitivity, specificity, and diagnostic odds ratio (DOR) were 0.81 (95% CI, 0.80-0.82), 0.64 (95% CI, 0.64-0.65), and 9.15 (95% CI, 5.72-14.62), respectively. The pooled estimates of the positive and negative likelihood ratios were 2.18 (95% CI, 1.66-2.87) and 0.28 (95% CI, 0.20-0.39), respectively. The area under the summary receiver operating characteristic (SROC) curve showed a discriminative performance of 0.77.
CONCLUSION: Deep learning based on retinal images had moderate diagnostic performance for detection of CKD. Relatively high sensitivity and modest specificity suggesting that deep learning is more effective for ruling out CKD than for confirming the diagnosis that deep learning is more effective for ruling out CKD than for confirming the diagnosis. Deep learning-based retinal analysis may therefore serve as a supportive, noninvasive screening or triage tool but should not replace established diagnostic methods for CKD.